Software Engineering / Data Systems

I turn data into decisions.

Software engineer and data scientist building tools that work for farmers, lenders, and communities where reliable tech matters most.

Secure Systems

Building protected apps that keep data safe from hackers.

Reliable Apps

Ensuring services work even in areas with weak connectivity.

Data Insights

Using math and statistics to find valuable business trends.

Conversational Analytics

Letting stakeholders ask questions of their data in plain English using AI agents and MCP servers.

Selected Systems

Credit Risk Assessment Engine

A new way for banks to score credit safely in emerging markets. This system uses data analysis to predict loan risk with 88% accuracy, protected by a secure API perimeter. It includes a Model Context Protocol (MCP) server that empowers risk officers to audit lending decisions and query portfolio distributions instantly using natural language AI agents.

The Problem:

Traditional credit scoring ignores profitable small businesses just because they don't have formal bank records, making it unfairly hard for them to get loans.

The Solution:

An AI engine that safely calculates loan risk by looking at everyday behaviors—like how regularly a business uses mobile money or pays utilities—instead of relying on old bank systems.

Business Value Created

01. Expanded Market

Safely underwrite thousands of previously invisible high-quality applicants.

02. Instant Execution

Sub-120ms automated decisions eliminate expensive manual underwriting delays.

03. Audit Readability

Automated SHAP trails guarantee regulatory compliance on every single loan decision.

04. Risk Protection

CAPIE-style defensive perimeters actively block adversarial probing attacks by default.

Live Demo Source Code Stakeholder Report

Execution Stack

FastAPI Docker LightGBM Backtesting MCP Server

Agri-Demand Optimizer

A predictive tool built to reduce post-harvest food waste in agricultural supply chains by forecasting buyer orders with 91% accuracy. It features a conversational data engine powered by an MCP server, allowing supply chain managers to bypass static dashboards and ask plain English questions to identify high-demand crops or estimate surplus reduction.

The Problem:

Farmers guess what to plant while commercial buyers order conservatively to avoid spoilage. This coordination failure results in up to 40% post-harvest food waste.

The Solution:

A delta-modeling pipeline separating purchase probability from volume regression, turning volatile agricultural data into stable 14-day advance buyer forecasts.

Business Value Created

01. Stable Farmer Income

Helping farmers know exactly how much to harvest to avoid losing money.

02. Accurate Ordering

Ensuring large kitchens always have the right amount of fresh produce.

03. Waste Reduction

Diverting tons of surplus food to where it's needed most.

04. Smarter Logistics

Planning better delivery routes to save time and fuel costs.

Live Demo Source Code Stakeholder Report

Supply Chain Stack

Demand Analytics LFS Deployment Delta-Modeling Kenyan Agritech MCP Server